A fundamental problem in programming languages is that (mostly) they expect you to sequence a number of events in time.
Functional and OO are not a big change over imperative in terms of that. Compilers and CPUs can reorder events and discover a little bit of parallelism but it's fundamentally limiting.
Declarative programming is something really different because there you are specifying what is to be done but not the order of events. The trouble with it is so far it has always been specialized (SQL) and there hasn't been a general toolbox to make it easy to make smart DSLs that really that efficient.
I believe that if you're looking for performance, you're never going to be able to avoid understanding how the programming language constructs are translated into actual machine instructions.
However, it is clearly less important these days to optimise code on that level, so as long as your runtime performs well enough, the benefit of having a simpler way to describe the solution is worth it.
Many would probably take issue with the statement above, but the popularity of Python, which is possibly the most slowest popular language that has ever been constructed, suggests that programmer efficiency is much more important to most people than performance.
Functional and OO are not a big change over imperative in terms of that. Compilers and CPUs can reorder events and discover a little bit of parallelism but it's fundamentally limiting.
Declarative programming is something really different because there you are specifying what is to be done but not the order of events. The trouble with it is so far it has always been specialized (SQL) and there hasn't been a general toolbox to make it easy to make smart DSLs that really that efficient.